A fan of Tesla might think that the automaker just can't catch a break when it comes to its autonomous driving tech. It’s already subject to several federal investigations over its marketing and deployment of technologies like Autopilot and Full Self-Driving (FSD), and as of last week, we can add another to the list involving around 2.4 million Tesla vehicles. This time, regulators are assessing the cars' performance in low-visibility conditions after four documented accidents, one of which resulted in a fatality.
The National Highway Traffic Safety Administration (NHTSA) says this new probe is looking at instances when FSD was engaged when it was foggy or a lot of dust was in the air, or even when glare from the sun blinded the car’s cameras and this caused a problem.
What the car can "see" is the big issue here. It's also what Tesla bet its future on.
Unlike the vast majority of its competitors that are giving their cars with autonomous driving capabilities more ways to “see” their surroundings, Tesla removed ultrasonic and other types of sensors in favor of a camera-only approach in 2022.
This means there isn’t really any redundancy in the system, so if a Tesla with FSD enabled drives through dense fog, it may not have an easy time keeping track of where the road is and staying on it. Vehicles that not only have cameras but also radar and lidar will make more sense of their environment even through dense fog, although these systems are also affected by the elements. Inclement weather seems to sometimes make FSD go rogue.
I didn’t realize they were using other sensors in the past and dropped them on newer models.
Older Teslas had a combination of radar and cameras for Autopilot and driver assistance systems. With newer software versions launched after Tesla went down the "Pure Vision" route, it disabled the sensors in the older cars that had them from the factory. So even if you have FSD enabled in an older Tesla that has more than just cameras, only the cameras will be used when the car is driving itself.
🤦♂️
Didn’t want to develop two different versions of software I guess?
Seriously though, wtf is up with Elon not liking LIDAR? I think pretty much every other manufacturer incorporates it into their higher-end driver assist stuff at this point.
Second, it would cost more money to put multiple types of sensors on the car. Spending money bad!
Personal speculation based on Elon's past behavior follows:
Plus he wanted to focus on visual recognition stuff likely because it would have multiple possible income streams compared to a sensor that is just good at keeping a car from running into things. Focusing on the visible light spectrum means the possibilities for facial recognition, data collection by a fleet of Teslas, including the ones people bought, taking pictures, etc.
Basically he wanted to focus on the one thing that seemed more profitable and didn't want to spend money on that stupid thing that just kept the car from crashing.
He's probably stuck on his decision to cut on LIDARs and compensate it with machine learning on cam inputs alone. That doesn't bring him the edge he wanted. Still, he doubles down as he's not risking anything besides being proclaimed wrong with that decision.
It’s hilarious because every single time he speaks about some unique aspect of starship that goes against conventional rocketry wisdom, like “we don’t need flame trenches. You get more efficiency on flat ground”, we just have to wait a year or two and all of a sudden they’re adding back the thing they tried to do without (see tower 2 flame trench going in as we speak).
I know a lot of companies go with RADAR over LIDAR because of reliability issues. RADAR is much more reliable because you can do it solid state, where LIDAR either has moving parts or is subject to IR bleed. However solid state LIDAR is finally becoming a thing so LIDAR will start becoming more commonplace in the next few years.
Humans are bad drivers as well. Technology should try to do better than humans, not accept the limitations of humans. When Radar, lidar (and others - possibly including things not invented yet) exist we should use them to make cars safer.
Well, actually, that's kind of the problem. It actually does more than what real autopilot does already. Autopilot in a plane can't help the plane not hit moving objects, it's not context aware at all. It just flies a pre-programmed route and executes pre-programmed maneuvers. Literally the first release was already better than what autopilot really is.
Planes are only safe because there is never supposed to be anything else anywhere near them. Which makes autopilot super easy. Which is why planes have had it since long before we had any context aware machines.
Also, if "roadspace" was treated the same as "airspace", including the amount of training and practice pilots have, as well as "road traffic controllers" like air traffic controllers. Self driving would have had no trouble right from the get-go. Pre-programmed routes, and someone making sure there is a specified gratuitous minimum space between each vehicle. And any violation being immediately harshly reprimanded...
Aitopilot is relatively easy compared to self-driving, if anything, calling it autopilot was being under ambitious.
But the autopilot on an aircraft or ship is often just a cruise control, maintaining a constant heading, speed, and (for aircraft) altitude. The pilot or skipper remains 100% responsible for course changes and collision avoidance.